Optimising quality of service and productivity of toll traffic management is a complex and difficult task; toll supervisors are asked to reduce operation costs, and therefore they hire no more toll collectors than required by traffic predictions. If, in a certain time period, the traffic is higher by 10% than they estimated, then drivers will have to wait before they pay toll fares at the highway toll plaza. COFIROUTE, a French toll highway operator, has developed and tested several tools either to measure the waiting time at toll plazas or to determine the length of queues. First developments using magnetic loop detectors, video processing and voting stands were not accurate enough; also, they were difficult to calibrate, due to many deviations in the functioning of the system. Then another original method was developed, using video image processing and license plate number recognition. The system has been implemented on a toll plaza on the highway network, and after some improvements, is now operational. This paper gives a presentation of different methods tested on the same site, and the performances and advantages of the vision system. (A)
Abstract